AI is finding vulnerabilities faster than humans can fix them, Anthropic warns

Anthropic revealed that its Glasswing initiative, powered by the Claude Mythos Preview model, uncovered over 10,000 high‑severity vulnerability candidates in just one month, far outpacing the open‑source community’s ability to patch them. Only 97 flaws have been fixed so far, highlighting a widening gap between AI‑driven discovery and human‑driven remediation. Mythos is proving dramatically more capable than previous models, identifying critical issues like a certificate‑forgery flaw in WolfSSL and even detecting real‑world fraud attempts, but its power also raises concerns about misuse if such capabilities become widely accessible. The situation underscores a new cybersecurity paradox: AI can protect systems better than ever, yet it also exposes weaknesses faster than developers can respond.

Read the full story on TNW →